WO2011084636A2

WO2011084636A2 - Novel methodology for arrhythmia risk stratification by assessing qt interval instability

Info

Publication number: WO2011084636A2
Application number: PCT/US2010/060843
Authority: WO
Inventors: Natalia A. Trayanova; Xiaozhong Chen
Original assignee: The Johns Hopkins University
Priority date: 2009-12-16
Filing date: 2010-12-16
Publication date: 2011-07-14
Also published as: US9215987B2; WO2011084636A3; US20130102912A1

Abstract

A method of predicting ventricular arrhythmias includes receiving an electrical signal from a subject's heart for a plurality of heart beats, identifying characteristic intervals and heart beat durations of the electrical signal corresponding to each of the plurality of heart beats to provide a plurality of characteristic intervals with corresponding heart beat durations, representing dynamics of the plurality of characteristic intervals as a function of a plurality of preceding characteristic intervals and durations of corresponding heart beats over a chosen period time, assessing a stability of the function over the chosen period of time, and predicting ventricular arrhythmias based on detected instabilities in the dynamics of the characteristic intervals.

Description

2240- 293156

NOVEL METHODOLOGY FOR ARRHYTHMIA RISK STRATIFICATION BY ASSESSING QT INTERVAL INSTABILITY

CROSS-REFERENCE OF RELATED APPLICATION

[0001] This application claims priority to U.S. Provisional Application No.

61/286,986 filed December 16, 2009, the entire contents of which are hereby incorporated by reference.

[0002] This invention was made with Government support of Grant No. HL082729, awarded by the Department of Health and Human Service, The National Institutes of Health (NIH). The U.S. Government has certain rights in this invention.

BACKGROUND

1. Field of Invention

[0003] The field of the currently claimed embodiments of this invention relates to systems and methods of predicting risk of ventricular arrhythmias.

2. Discussion of Related Art

[0004] Sudden cardiac death (SCD) is a major health problem in the industrialized world. SCD often occurs in healthy individuals without prior history of heart disease. The early diagnosis of arrhythmia has the potential to significantly reduce mortality due to SCD. Clinical studies have demonstrated that instability in the QT interval of the electrocardiogram (ECG), which is the global manifestation of instability in ventricular repolarization, predicts propensity to arrhythmias. However, a robust methodology to assess QT interval instability and thus predict propensity to arrhythmia in a patient- specific manner is currently lacking. 2240- 293156

[0005] Clinical studies have demonstrated that the QT interval instability in the ECG signal is associated with a propensity for lethal arrhythmias. A methodology that allows the physician to predict the development of instabilities in the QT interval in a patient-specific manner, and thus predict the patient's propensity for arrhythmias, would be a significant step forward in improving SCD risk stratification and in preventing loss of human life.

[0006] ECG (or electrogram if recorded from an implanted cardioverter-defibrillator

(ICD) device) is the most widely-used tool in the diagnosis of heart rhythm disorders. The QT interval of the ECG (interval between Q and T deflections, Figure 1A) has been found to be unstable in the diseased heart (Figure IB). QT interval instability has been reported in long QT syndrome patients (Merri M, Moss AJ, Benhorin J, Locati EH, Alberti M, Badilini F. Relation between ventricular repolarization duration and cardiac cycle length during 24- hour Holier recordings. Findings in normal patients and patients with long QT syndrome. Circulation. 1992;85(5):1816-1821; Yamauchi S, Yamaki M, Watanabe T, Yuuki K, Kubota I, Tomoike H. Restitution properties and occurrence of ventricular arrhythmia in LQT2 type of long QT syndrome. / Cardiovasc Electrophysiol. 2002; 13(9):910-914; Chinushi M, Restivo M, Caref EB, El-Sherif N. Electrophysiological basis of arrhythmogenicity of QT/T alternans in the long-QT syndrome: tridimensional analysis of the kinetics of cardiac repolarization. Circ Res. 1998;83(6):614-628). QT prolongation and abnormal QT dynamics have been found in acute myocardial infarction patients (Chevalier P, Burri H, Adeleine P, Kirkorian G, Lopez M, Leizorovicz A, Andre-Fouet X, Chapon P, Rubel P, Touboul P. QT dynamicity and sudden death after myocardial infarction: results of a long-term follow-up study. / Cardiovasc Electrophysiol. 2003;14(3):227-233; Schwartz PJ, Wolf S. QT interval prolongation as predictor of sudden death in patients with myocardial infarction. Circulation. 1978;57(6): 1074-1077; Bonnemeier H, Wiegand UK, Bode F, Hartmann F, Kurowski V, Katus HA, Richardt G. Impact of infarct-related artery flow on QT dynamicity in patients undergoing direct percutaneous coronary intervention for acute myocardial infarction.

Circulation. 2003;108(24):2979-2986; Szydlo K, Trusz-Gluza M, Wita K, Filipecki A, Orszulak W, Urbanczyk D, Krauze J, Kolasa J, Tabor Z. QT/RR relationship in patients after remote anterior myocardial infarction with left ventricular dysfunction and different types of ventricular arrhythmias. Ann Noninvasive Electrocardiol. 2008; 13(l):61-66). Berger et al. 2240- 293156 reported an increased QT interval variability in dilated cardiomyopathy patients (Figure IB) (Berger RD, Kasper EK, Baughman KL, Marban E, Calkins H, Tomaselli GF. Beat-to-beat QT interval variability: novel evidence for repolarization lability in ischemic and

nonischemic dilated cardiomyopathy. Circulation. 1997;96(5): 1557- 1565). These studies indicate that the instability in the QT interval is an important indicator of cardiac rhythm disorder.

[0007] Instability in the QT interval is a manifestation of instability in repolarization in the heart. At the cellular level, instability in myocyte repolarization is assessed from the dynamics of its action potential duration (APD). APD instability and its implication for arrhythmogenesis have been extensively studied (Laurita KR, Girouard SD, Rosenbaum DS. Modulation of ventricular repolarization by a premature stimulus. Role of epicardial dispersion of repolarization kinetics demonstrated by optical mapping of the intact guinea pig heart. Circ Res. 1996;79(3):493-503; Akar FG, Rosenbaum DS. Transmural

electrophysiological heterogeneities underlying arrhythmogenesis in heart failure. Circ Res. 2003;93(7):638-645; Sampson KJ, Henriquez CS. Simulation and prediction of functional block in the presence of structural and ionic heterogeneity. Am J Physiol

2001;281(6):H2597-2603; Chen X, Fenton FH, Gray RA. Head-tail interactions in numerical simulations of reentry in a ring of cardiac tissue. Heart Rhythm. 2005;2(9): 1038-1046; Weiss JN, Chen PS, Qu Z, Karagueuzian HS, Lin SF, Garfinkel A. Electrical restitution and cardiac fibrillation. / Cardiovasc Electrophysiol. 2002;13(3):292-295; Gilmour RF, Chialvo DR. Electrical Restitution, Critical Mass, and the Riddle of Fibrillation. J. Cardiovasc

Electrophysiol. 1999; 10(8): 1087- 1089; Banville I, Gray RA. Effect of action potential duration and conduction velocity restitution and their spatial dispersion on alternans and the stability of arrhythmias. J Cardiovasc Electrophysiol. 2002;13(11): 1141-1149; Elharrar V, Surawicz B. Cycle length effect on restitution of action potential duration in dog cardiac fibers. Am J Physiol. 1983;244(6):H782-792; Gilmour RF, Otani NF, Watanabe MA.

Memory and complex dynamics in cardiac Purkinje fibers. Am J Physiol.

1997;272(4):H1826; Franz MR, Swerdlow CD, Liem LB, Schaefer J. Cycle length dependence of human action potential duration in vivo. Effects of single extrastimuli, sudden sustained rate acceleration and deceleration, and different steady- state frequencies. / Clin 2240- 293156

Invest. 1988;82(3):972-979). It is thus reasonable to expect that concepts developed to determine instability in APD could be translated to the clinic and applied in the evaluation of the patient's QT interval instability. The efforts to assess instability in APD are therefore reviewed below.

[0008] APD depends on the preceding diastolic interval (DI). This relationship is termed APD restitution. A standard APD restitution curve is constructed by delivering S2 pacing stimuli following the same S I pacing train for a broad range of S 1-S2 intervals; S I pacing is at a constant rate. In a dynamic APD restitution protocol, tissue is paced continuously at different rates; the last beat for each pacing episode is used to construct the APD restitution. The restitution hypothesis postulates that APD can be predicted from the preceding DI and the restitution relationship (Figure 2), and implicates the maximum slope of the APD restitution curve as the sole predictor of repolarization instability and thus of propensity to arrhythmia (Nolasco JB, Dahlen RW. A graphic method for the study of alternation in cardiac action potentials. JAppl Physiol. 1968;25(2): 191-196; Garfinkel A, Kim YH, Voroshilovsky O, Qu Z, Kil. Preventing ventricular fibrillation by flattening cardiac restitution. PNAS. 2000;97(11); Riccio ML, Koller ML, Gilmour RF. Electrical restitution and spatiotemporal organization during ventricular fibrillation. Circ Res.

1999;84:955-963). Indeed, studies have shown that steep restitution increases the functional heterogeneity of repolarization (Garfinkel A, Kim YH, Voroshilovsky O, Qu Z, Kil.

Preventing ventricular fibrillation by flattening cardiac restitution, PNAS, 2000;97(11);

Riccio ML, Koller ML, Gilmour RF. Electrical restitution and spatiotemporal organization during ventricular fibrillation. Circ Res, 1999;84(955-963); Qu Z, Garfinkel A, Chen PS, Weiss JN. Mechanisms of discordant alternans and induction of reentry in simulated cardiac tissue. Circulation. 2000;102(14): 1664-1670; Cao JM, Qu Z, Kim YH, Wu TJ, Garfinkel A, Weiss JN, Karagueuzian HS, Chen PS. Spatiotemporal heterogeneity in the induction of ventricular fibrillation by rapid pacing: importance of cardiac restitution properties. Circ Res. 1999;84(11): 1318-1331), and leads to arrhythmogenesis (Akar FG, Rosenbaum DS.

Transmural electrophysiological heterogeneities underlying arrhythmogenesis in heart failure. Circ Res. 2003;93(7):638-645; Chen X, Fenton FH, Gray RA. Head-tail interactions in numerical simulations of reentry in a ring of cardiac tissue. Heart Rhythm 2005;2(9): 1038- 2240- 293156

1046). Studies have also reported that a steep restitution causes APD alternans at the cellular level (Nolasco JB, Dahlen RW. A graphic method for the study of alternation in cardiac action potentials. J Applied Physiol. 1968;25(2): 191-196) and results in spiral wave breakup and the induction of ventricular fibrillation (VF) (Weiss JN, Chen PS, Qu Z, Karagueuzian HS, Lin SF, Garfinkel A. Electrical restitution and cardiac fibrillation. / Cardiovasc

Electrophysiol. 2002;13(3):292-295; Gilmour RF, Chialvo DR. Electrical Restitution, Critical Mass, and the Riddle of Fibrillation. J. Cardiovasc Electrophysiol. 1999; 10(8): 1087- 1089).

[0009] The assessment of APD stability, and thus the prediction of propensity to arrhythmia based on the criterion of maximum restitution slope > 1 is, however, not always accurate: arrhythmia has been induced with slope < 1, and failed to be induced with slope > 1 (Elharrar V, Surawicz B. Cycle length effect on restitution of action potential duration in dog cardiac fibers. Am J Physiol. 1983;244(6):H782- 792; Gilmour RF, Otani NF, Watanabe MA. Memory and complex dynamics in cardiac Purkinje fibers. Am J Physiol.

1997;272(4):H1826; Franz MR, Swerdlow CD, Liem LB, Schaefer J. Cycle length dependence of human action potential duration in vivo. Effects of single extrastimuli, sudden sustained rate acceleration and deceleration, and different steady- state frequencies. / Clin Invest. 1988;82(3):972-979; Saitoh H, Bailey JC, Surawicz B. Action potential duration alternans in dog Purkinje and ventricular muscle fibers. Further evidence in support of two different mechanisms. Circulation. 1989;80(5): 1421- 1431; Karagueuzian HS, Khan SS, Hong K, Kobayashi Y, Denton T, Mandel WJ, Diamond GA. Action potential alternans and irregular dynamics in quinidine-intoxicated ventricular muscle cells. Implications for ventricular proarrhythmia. Circulation. 1993;87(5): 1661-1672). Huang et al., evaluating the effect of previous activations on APD with a statistical approach (Huang J, Zhou X, Smith WM, Ideker RE. Restitution properties during ventricular fibrillation in the in situ swine heart. Circulation. 2004;110(20):3161-3167), reported that during VF, the contribution of preceding APD to current APD is as significant as the contribution of the preceding DI. These results indicate that events that precede DI also affect stability of APD.

[0010] The events that precede DI represent the activation history, i.e. memory

(Figure 3). Memory has been found to affect APD and its restitution. Elharrar et al (Elharrar 2240- 293156

V, Surawicz B. Cycle length effect on restitution of action potential duration in dog cardiac fibers. Am J Physiol. 1983;244(6):H782-792) reported that APD restitution depended on pacing rate. Gilmour and Otani (Otani NF, Gilmour RF. Memory Models for the Electrical Properties of Local Cardiac Systems. J. Theor. Biol. 1997;187:409-436) found that that memory can enhance or diminish APD stability. These studies indicate that both restitution and memory govern APD stability.

[0011] Similar to instability in APD, the restitution concept has been used to evaluate

QT interval instability in the clinic by evaluating the slope of the dependence of the QT interval on the preceding TQ interval or preceding RR interval (since QT interval + QT interval is equal to RR interval) (Figure 1 A). The latter relationship has been tested as an arrhythmia risk predictor (Yamauchi S, Yamaki M, Watanabe T, Yuuki K, Kubota I, Tomoike H. Restitution properties and occurrence of ventricular arrhythmia in LQT2 type of long QT syndrome. / Cardiovasc Electrophysiol. 2002; 13(9):910-914; Fossa AA,

Wisialowski T, Crimin K, Wolfgang E, Couderc J-P, Hinterseer M, Kaab S, Zareba W, Badilini F, Sarapa N. Analyses of Dynamic Beat-to-Beat QT-TQ Interval (ECG Restitution) Changes in Humans under Normal Sinus Rhythm and Prior to an Event of Torsades de Pointes during QT Prolongation Caused by Sotalol. Ann Noninvasive Electrocardiol.

2007;12(4):338 - 348; Gilmour RF, Riccio ML, Locati EH, Maison-Blanche P, Coumel P, Schwartz PJ. Time- and rate-dependent alterations of the QT interval precede the onset of torsade de pointes in patients with acquired QT prolongation. J Am Coll Cardiol.

1997;30:209-217). However, such an approach is invasive and requires constant pacing, thus eliminating the role of memory (i.e. the dependence of a QT interval on preceding TQ and QT intervals). It is therefore clear that a robust methodology, which can detect instability of the QT interval directly from the clinical ECG and without invasive pacing is currently lacking. If such a methodology is developed, it could be used to predict an impending arrhythmia event from a patient clinical ECG or from an ICD electrogram recording. 2240- 293156

SUMMARY

[0012] A method of predicting ventricular arrhythmias according to an embodiment of the current invention includes receiving an electrical signal from a subject's heart for a plurality of heart beats, identifying characteristic intervals and heart beat durations of the electrical signal corresponding to each of the plurality of heart beats to provide a plurality of characteristic intervals with corresponding heart beat durations, representing dynamics of the plurality of characteristic intervals as a function of a plurality of preceding characteristic intervals and durations of corresponding heart beats over a chosen period time, assessing a stability of the function over the chosen period of time, and predicting ventricular arrhythmias based on detected instabilities in the dynamics of the characteristic intervals.

[0013] A medical device according to an embodiment of the current invention has a system for predicting ventricular arrhythmias. The system has a data processor configured to receive an electrical signal from a subject's heart for a plurality of heart beats, identify characteristic intervals and heart beat durations of the electrical signal corresponding to each of the plurality of heart beats to provide a plurality of characteristic intervals with corresponding heart beat durations, represent the dynamics of the plurality of characteristic intervals as a function of a plurality of preceding characteristic intervals and durations of the corresponding heart beats over a chosen period of time, assess a stability of the function over the chosen period of time, and predict ventricular arrhythmias based on detected instabilities in the dynamics of the characteristic intervals.

[0014] A computer readable medium according to an embodiment of the current invention has stored executable instructions for execution by a computer. The executable instructions include executable instructions for receiving an electrical signal from a subject's heart for a plurality of heart beats, identifying characteristic intervals and heart beat durations of the electrical signal corresponding to each of the plurality of heart beats to provide a plurality of characteristic intervals with corresponding heart beat durations, representing dynamics of the plurality of characteristic intervals as a function of a plurality of preceding characteristic intervals and durations of corresponding heart beats over a chosen period time, 2240- 293156 assessing a stability of the function over the chosen period of time, and predicting ventricular arrhythmias based on detected instabilities in the dynamics of the characteristic intervals.

BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Further objectives and advantages will become apparent from a consideration of the description, drawings, and examples.

[0016] Figure 1A shows an ECG trace with phases and intervals.

[0017] Figure IB shows an example of unstable QT dynamics recorded from a dilated cardiomyopathy patient.

[0018] Figure 2 a schematic illustration indicating that restitution is the dependence of APD on the preceding DI.

[0019] Figure 3 is a schematic illustration of APD as a function of the preceding DI and memory. Here memory is an independent variable.

[0020] Figure 4 is a schematic illustration of a system and method according to an embodiment of the current invention. In an embodiment, the QT interval (QTI) is a function of the preceding TQ interval (TQI), or equivalently of the preceding RR interval (RRI) since RRI=QTI+TQI, and memory. Memory is represented by the QTIs and RRIs (or TQIs) of M previous beats.

[0021] Figure 5 illustrates ECG labeling used for an example according to the current invention.

[0022] Figure 6A is an epicardial view of the image-based human ventricles model

(ventricles in light, atria in dark). Atria were isolated from the ventricles during the simulation. ECG electrodes (El and E2) and pacing electrode (E3) are marked.

[0023] Figure 6B is a transmural view of the model. 2240- 293156

[0024] Figure 6C shows action potential morphology of Endo-, M, and Epicardial cells.

[0025] Figure 6D shows RRI sequence with the artificially induced PAs. The DOP of PA was 0ms, 150ms, 200ms, or 220ms respectively at each pacing train.

[0026] Figures 7A-7B show predicted QTI (QTI_p) dynamics of a minECG, compared with the same dynamics extracted directly from the patient ECG recording for M_max =24 (Figure 7 A) and M=3 (Figure 7B).

[0027] Figure 7C shows the dependence of prediction accuracy on the value of M for the example of Figures 7A-7B.

[0028] Figure 8A shows a comparison of N_us between the VT group and the control group.

[0029] Figure 8B shows a comparison of fpA between the VT group and the control group.

[0030] Figure 9 shows the QTI restitution constructed from a minECG before the onset of VT according to an embodiment of the current invention. PA was presented in this minECG. Exponential QTI restitution curves were separately fitted in ranges TQI<=400ms and TQI>400ms.

[0031] Figure 10 shows the relationship between the sensitivity and 1-specificity of the prediction at different QTII values. The area under the curve indicates the accuracy of the prediction. The larger the area, the more accurate the prediction.

[0032] Figure 11 shows the dynamics of survival probability in time for patients identified as being of either high risk (QTII= >0.9) or low risk (QTII < 0.9) of arrhythmia. The difference between the two groups is statistically significant (p=0.001). 2240- 293156

DETAILED DESCRIPTION

[0033] Some embodiments of the current invention are discussed in detail below. In describing embodiments, specific terminology is employed for the sake of clarity. However, the invention is not intended to be limited to the specific terminology so selected. A person skilled in the relevant art will recognize that other equivalent components can be employed and other methods developed without departing from the broad concepts of the current invention. All references cited anywhere in this specification, including the Background and Detailed Description sections, are incorporated by reference as if each had been individually incorporated.

[0034] In view of the above-summarized background, we considered the following question: Could the knowledge regarding the fact that APD instability is a function of both restitution and memory be translated to the clinic and used to detect patient QT interval instability and thus, his/her propensity to arrhythmia, directly from the clinical ECG and without invasive pacing? As mentioned above, the restitution concept has indeed been used to evaluate QT interval instability, however, such an approach is invasive and requires constant pacing, thus eliminating the role of memory

[0035] The current approaches for assessing QT interval stability are statistical. Most notable among them is the QT variability index (QTVI) (Berger et al., supra; Murabayashi T, Fetics B, Kass D, Nevo E, Gramatikov B, Berger RD. Beat-to-beat QT interval variability associated with acute myocardial ischemia. J Electrocardiol. 2002;35(l): 19-25; Vrtovec B, Stare V, Stare R. Beat-to-beat QT interval variability in coronary patients. J Electrocardiol. 2000;33(2): 119-125). Such approaches evaluate the mean and standard deviation of the QT interval duration over the entire ECG, and thus provide an overall estimation of QT interval instability. However, they do not capture the development of QT interval instability from one beat to the next. Furthermore, ectopic beats are typically excluded from QTVI analysis, thus the approach focuses predominantly on the contribution of the autonomic nervous system (ANS) to QT interval stability. This contribution is, however, diminished in the diseased heart (Berger RD, Kasper EK, Baughman KL, Marban E, Calkins H, Tomaselli GF. Beat-to-beat QT interval variability: novel evidence for repolarization lability in ischemic and nonischemic dilated cardiomyopathy. Circulation. 1997;96(5): 1557- 1565; Kleiger R, 2240- 293156

Miller J, Bigger JJ, Moss A. Decreased heart rate variability and its association with increased mortality after acute myocardial infarction. Am J Cardiol. 1987;59(4):256-262).

[0036] It is clear that a robust methodology, which can detect instability of the QT interval directly from the clinical ECG and without invasive pacing is currently lacking in the prior art. If such a methodology is developed, it could be used to predict an impending arrhythmia event from a patient clinical ECG or from an ICD electrogram. Some

embodiments of the current invention can fill this void. Some embodiments of the current invention use, as a basis, the concepts and approaches to APD stability reviewed above. We then radically modify and extend these approaches to make them capable of assessing QT interval stability from clinical ECG recordings. All factors contributing to dynamic instability in QT interval, such as ANS influences as well as ectopic beats (the latter unmask QT interval instability in a manner much like small DIs unmask large APD restitution slopes) can be accounted for in embodiments of the current invention, rendering it a powerful tool in the assessment of propensity to arrhythmia from clinical ECG or ICD electrogram

recordings. Although various embodiments of the current invention have utility in clinical practice, other embodiments of the current invention can also be used in basic science studies to evaluate APD stability.

[0037] Some embodiments of the current invention provide methods for detecting QT interval instability directly from clinical ECG or ICD electrogram recordings. Embodiments of the current invention provide a novel approach that is based on engineering methods never previously applied to cardiac electrophysiology. Application of this methodology in the clinic can provide methods and devices to monitor and detect QT interval instability and allow the clinician to assess the patient's propensity to arrhythmia, for example. Prediction of arrhythmia onset could provide a vital time window for early intervention and save human life. In the current environment which emphasizes reducing health care costs and optimizing therapy, robust diagnostic approaches to identify patients most susceptible to SCD would have a dramatic personal, medical and economic impact on the lives of many individuals.

[0038] Some aspects of the current invention provide a novel and radically different approach that overcomes the limitations of previous approaches that have failed to translate 2240- 293156 basic science concepts of restitution and cardiac memory into clinical applications. Systems and methods according to some embodiments of the current invention could be implemented in Holier monitors, treadmill stress tests, ICDs, and ECG-based devices, including those that monitor patients at locations remote from the hospital. Although some embodiments of the current invention are useful in clinical practice, some embodiments can be applied with equal ease in experiments and simulations, and at the level of the cell or tissue to assess local APD stability, and thus to facilitate the mechanistic inquiry into lethal ventricular

arrhythmogenesis .

[0039] Figure 4 is a schematic illustration to facilitate the description of a method and system for predicting ventricular arrhythmias according to an embodiment of the current invention. The method of predicting ventricular arrhythmias according to this embodiment of the current invention includes receiving an electrical signal from a subject's heart for a plurality of heart beats, identifying characteristic intervals and heart beat durations of the electrical signal corresponding to each of the plurality of heart beats to provide a plurality of characteristic intervals with corresponding heart beat durations, representing dynamics of the plurality of characteristic intervals as a function of a plurality of preceding characteristic intervals and durations of corresponding heart beats over a chosen period time (which can be as short as a single heart beat or longer), assessing the stability of this function (i.e. the stability of the dynamics of the characteristic intervals) over the chosen period of time, and predicting ventricular arrhythmias based on detected instabilities in the dynamics of the characteristic intervals over the said period of time. The electrical signal received can be an ECG or ICD electrogram signal, for example. However, other types of measurements of electrical activity of a subject's heart could also be used in other embodiments of the current invention. The characteristic interval can be a QT interval of an ECG or ICD electrogram signal according to some embodiments of the current invention.

[0040] The modeling of the plurality of characteristic intervals can include representing each QT interval as a function of a plurality of QT intervals and RR (or TQ) intervals. The modeling of the plurality of characteristic intervals can be done, but is not limited to, employing a linear autoregressive model with exogenous inputs (ARX) model, for example. The determining of the parameters of the QT interval dynamics model, according 2240- 293156 to an embodiment of the current invention based on the plurality of QT and RR (or TQ) intervals, can be done by determining the parameters a; and b; in the equation

M M

Q?I_n =∑ a, x QTI^ +∑ b_t x RRI^

i=l i=l where QTI and RRI are QT and RR intervals, respectively; M is a number of heart beats included in the determining (this is memory); and n is a heart beat number.

[0041] The determining the instability in the dynamics of the characteristic intervals can include determining the instability of the ARX model above according to an embodiment of the current invention. The determining the instability of the ARX model can include taking a z-transform of the ARX model equation according to some embodiments of the current invention.

[0042] The method of predicting ventricular arrhythmias according to some embodiments of the current invention can also include evaluating the dependence of the instability in the dynamics of the characteristic intervals on the presence of premature heart beats. According to some embodiments, this can be done by counting the number of premature heart beats in the same periods of time over which instabilities in the dynamics of the characteristic intervals are evaluated. The method can further include calculating a ratio of the number of time periods with unstable dynamics in the ECG or ICD electrogram, divided by the number of time periods with premature heart beats, to provide a stratification index, such that predicting ventricular arrhythmias is based on the stratification index according to some embodiments of the current invention.

[0043] Figure 4 schematically represents a medical device according to some embodiments of the current invention, which includes a system for predicting ventricular arrhythmias. The system includes a data processor configured to receive an electrical signal from a subject's heart for a plurality of heart beats, identifying characteristic intervals and heart beat durations of the electrical signal corresponding to each of the plurality of heart beats to provide a plurality of characteristic intervals with corresponding heart beat durations, representing dynamics of the plurality of characteristic intervals as a function of a plurality of 2240- 293156 preceding characteristic intervals and durations of corresponding heart beats over a chosen period time (which can be as short as a single heart beat or longer), assessing the stability of this function (i.e. the stability of the dynamics of the characteristic intervals) over the chosen period of time, and predicting ventricular arrhythmias based on detected instabilities in the dynamics of the characteristic intervals. The medical device can be, but is not limited to, a Holier monitor, a treadmill stress test device, an implantable cardioverter-defibrillators, or an ECG-based device, including devices that can monitor and/or respond in accordance with predictions of coronary arrhythmias according to an embodiment of the current invention.

[0044] Further embodiments of the current invention include a computer readable medium that includes stored executable instructions for execution by a computer. The executable instructions include instructions for receiving an electrical signal from a subject's heart for a plurality of heart beats, identifying characteristic intervals and heart beat durations of the electrical signal corresponding to each of the plurality of heart beats to provide a plurality of characteristic intervals with corresponding heart beat durations, representing dynamics of the plurality of characteristic intervals as a function of a plurality of preceding characteristic intervals and durations of corresponding heart beats over a chosen period time (which can be as short as a single heart beat or longer), assessing the stability of this function (i.e. the stability of the dynamics of the characteristic intervals) over the chosen period of time, and predicting ventricular arrhythmias based on detected instabilities in the dynamics of the characteristic intervals..

[0045] Further embodiments of the current invention include computer programs configured to perform the methods of the current invention.

EXAMPLE 1

[0046] Instability in the dynamics of ventricular repolarization plays an important role in the mechanisms of arrhythmia, especially when a premature heart beat (i.e. premature activation, PA) is presented. At the cellular level, ventricular repolarization is measured by the action potential duration (APD), the latter often expressed as a function of the preceding diastolic interval (DI). This relationship is known as APD restitution. Over the past decade, much emphasis has been placed on the restitution curve slope as a major factor in both the 2240- 293156 onset of arrhythmias, and the dynamic destabilization of reentrant waves that underlie arrhythmias. A large (>1) APD restitution slope indicates instability in APD dynamics; 7-^"9 it is observed in the short DI range, which is typically associated with PAs. The shorter the DI, the larger the restitution slope. Research has demonstrated that unstable APD dynamics results in spatial gradients in APD, 10-^"12 leads to ventricular tachycardia (VT) following PA, 1-^"3 and causes the transition from VT to ventricular fibrillation (VF). 8 ' 9 ' 12

[0047] To test the importance of APD restitution slope as an arrhythmia risk index in the clinic, several studies have measured this slope in patients using monophasic action potential (MAP) catheters.¹³' ¹⁴ These studies have reported increased APD restitution slopes in diseased human hearts, as identified by invasive constant-rate pacing protocols. The use of such pacing protocols eliminated short-term memory, which is the dependence of APD, in addition to the preceding DI, on activation history, typically represented by several previous APDs and DIs.¹⁵' ¹⁶ Since short-term memory can either enhance or diminish the instability in APD dynamics, 17 ' 18 APD restitution slope cannot be used as an accurate measure of APD instability.

[0048] The QT interval (QTI) in an ECG is the global manifestation of ventricular

APD. Based on the concept of APD restitution, larger-than-one slope of the QTI restitution (the dependence of QTI on the preceding TQ interval, TQI), which is an indicator of instability in QTI dynamics, has been tested as an index of arrhythmia risk. 17 ' 19 ' 20 However, similar to APD restitution slope, QTI restitution slope has to be measured under invasive constant-rate pacing protocols, which similarly eliminate the effect of short-term memory on QTI instability. Currently, there is no robust methodology that can detect instability in QTI non-invasively from the clinical ECG and without canceling the important contribution of short-term memory. Furthermore, it remains unknown whether instability in QTI dynamics, when detected, would be predictive of arrhythmia risk. Finally, the contributions of PA to instability in QTI dynamics and the onset of arrhythmias are also unknown. It these relationships are determined, a risk stratification index to predict the arrhythmia risk based on QTI instability can be developed. 2240- 293156

[0049] In this example according to an embodiment of the current invention, we provide a novel methodology for detecting instability in QTI dynamics. The algorithm defines QTI as a function of both its preceding RR interval (RRI; RRI=QTI + TQI) and activation history (several preceding QTI and RRIs), and assesses the instability of this function. Because the effect of short-term memory is included in the algorithm, the resulting indicator of instability is more accurate than restitution slope. The methodology uses the clinical ECG and does not require a specific pacing protocol; it is thus noninvasive. In this example we test the hypotheses that instability in QTI dynamics, as detected by this methodology, precedes VT onset. Furthermore, we test the additional hypothesis that instability in QTI dynamics is positively correlated with the PA's degree of prematurity (DOP). The results of this example reveal the important contribution of PA and QTI instability to VT onset. Confirming these this hypothesies indicates that the present methodology according to some embodiments of the current invention for detecting instability in QTI dynamics could be used to develop an index that predicts risk of arrhythmias.

[0050] We tested the first hypothesis using clinical ECG recordings collected from acute myocardial infarction (AMI) patients. However, an ECG recording may have multiple PAs, each with different DOP, which makes directly testing the second hypothesis difficult. Instead, we used pseudo-ECG recordings generated using an MRI-based model of human ventricles, so that the DOP of PA can be controlled.

Methods

[0051] The details of the clinical data collection, the computer simulations, and the algorithm for detecting instability in QTI dynamics according to an embodiment of the current invention are presented in the following sections.

ECG data

[0052] The ECG recordings used in this study were part of a clinical ECG database at the Johns Hopkins Hospital. All ECGs in the database were recorded with the specialized intensive care unit MARS telemetry system (GE Medical Systems, Milwaukee, WI). This 2240- 293156 system continuously records up to 28 hours of multi-lead ECG, sampled at 125 Hz. From this database, 15 patients with either sustained (10) or non-sustained (5) monomorphic VT events were chosen. Among the 15 chosen patients, 7 (46.7%) were male, 2 (13.3%) were African American, and 13 (86.7%) were white. The mean average age was 67.2+3.9 years. All patients had coronary artery disease (CAD) and 6 (40%) patients used beta-blocker. From the multi-lead recordings of each patient, the recording with the best signal-to-noise ratio was chosen for analysis, thus avoiding additional filtering. VT events were identified from these ECG recordings by the cardiologist. Ten-minute-long ECG recordings, extracted immediately before VT onset and at least 1 hour before or after any arrhythmia event, were assembled into a VT and a control group, respectively, with one trace per patient in each group. Each 10-minute trace was then divided into ten 1 -minute ECG recordings (minECGs).

[0053] For each minECG, Q, R, and T waves were labeled to extract QTI, TQI, and

RRI. The QRS complex was first identified, followed by identification of the T wave and the isoelectric line using the approach of Laguna et al. 22 The end of the T wave was then found as the intersection of the isoelectric line with the tangent to the T wave at the point of maximum slope, as illustrated in Figure 5. QTI, TQI, and RRI were defined as shown in Figure 5. All labeling was visually checked to ensure accuracy. The number of PAs was also counted for each patient. Following the approach of Huikuri et al., 23 a PA was counted when the RRI of the beat was shortened by at least 100ms with respect to that of the preceding beat.

Computer simulations

[0054] We used the newly-developed anatomically realistic model of the human ventricles (Figure 6A), reconstructed from MRI and diffusion-tensor (DT) MRI data, 24 to examine the effect of the DOP of PA on QTI instability. In this study, the DOP of a beat was defined as the shortening of the CI with respect to the one of the preceding beat.

[0055] Description of our pipeline for the model construction was published previously. 24 ' 25 Mathematical description of the electrical behavior of cardiac tissue was based on the monodomain representation, ensuring speed of execution of the simulations. 2240- 293156

The membrane dynamics of the human ventricular cell was represented by the ten Tuscher et al ionic model.²⁶ In order to obtain an ECG consistent with the clinical signal, we included transmural heterogeneities in cellular properties²⁶ in the model ventricles. The spatial distribution of endo- epi-, and M-cells across the ventricular walls (Figure 6C) was implemented as in a previous study; 27 it was based on data by Drouin et al.28

[0056] A one-minute long pacing train was delivered epicardially at the apex of the human ventricular model (Figure 6A). The pacing intervals in this train were the same as the RRIs of one of the PA-free minECGs in the control group. The timings of 4 randomly selected beats of this train were shortened to simulate PAs (Figure 6D), with all 4 beats assumed to have the same DOP. This pacing protocol was implemented four times, with the same PA timings but different DOP values, namely 0ms, 150ms, 200ms, and 220ms. During each pacing protocol, one-minute long pseudo-ECGs (pseudo-minECGs) were computed as done previously. 27 The ECG leads were configured to record the Lead II of a standard 12- lead ECG system (Figure 6A).

Assessment of instability in QTI dynamics

[0057] For each minECG or pseudo-minECG, the dependence of a given QTI on the preceding QTIs and RRIs was expressed as Equation 1, which is a linear autoregressive model with exogenous input (ARX). The latter has been used previously to describe APD dynamics.¹⁶' ²⁹

M M

QTI_n =∑a_i x QTI_n__i +∑b_i x RRI_n__i Equation 1 i=l i=l

[0058] In Equation 1, n is the beat number in the minECG or pseudo-minECG; <¾· and bi (i=l,..,M) are the weights (constants) with which each preceding QTI and RRI, respectively, contributes to QTI_n. M is the number of beats representing the extent of the activation history.

[0059] The parameters of each ARX model were evaluated with an established approach termed Steiglitz-McBride iteration. 31 For each ARX model (i.e. each minECG or 2240- 293156 pseudo-minECG), the value of M was determined by increasing it from 1, in steps of 1, and examining, at each step, whether an accurate prediction of QTI dynamics in the minECG was achieved. The value of M at which the prediction reaches a predetermined accuracy is denoted as M_max. The predetermined accuracy in this study was that the mean square error between the predicted and the measured QTI was smaller than 5 ms .

[0060] We used a well-established method³² to assess the stability of the ARX model

(i.e. of Equation 1) and expressed it as a stability index, P_m. The definition and calculation of P_m follows, below. Based on the value of P_m, the ARX model was found stable (P_m <1) or unstable (P_m >=1). 32 When the ARX model was unstable, the degree of instability was manifested in the value of P_m; the larger the value of P_m, the larger the instability.

[0061] To assess the stability of the ARX model, Equation 1 was rewritten as:

M M

∑α,· ^χ βΓ„_. = ¾ x RR^ Equation 2 i=0 i=l

We applied a z-transform to Equation 2:

QT(zia₀ + α,ζ^'1 + a₂z^~2 + .... + a_M z^~M ) = RRiztb.z^'1 + b₂z^~2 + .... + b_Mz^~M ). Equation 3 [0062] and re- wrote the resulting equation as:

QT{z) ₌ b_lZ-¹ + b₂z-² + .... + b_Mz-^M

RR{z) a₀ + a, z^_1 + a₂z^~2 + .... + a_Mz^~M , , . .

M Equation 4

[0063] Equation 4 is the transfer function of the ARX model expressed as Equation

1. Each a (or β) in Equation 4 is a complex number, and is called a pole (or a zero). The value of each pole (or zero) is determined by <¾ (or bi) (i=0 to M), which are the weights determining the contribution of each QTI (or RRI). A pole can be cancelled by a zero if they are equal to each other. It can be proved that a system defined by Equation 4 is bounded- input bounded-output (BIBO) unstable if the magnitude of any un-cancelled pole is bigger 2240- 293156 than 1. In this example, the maximum magnitude of all un-cancelled poles is termed as QTI instability index P_m. The z-transform operation and the calculation of pole and zero can be done with Matlab functions.

Data Analysis

[0064] Using the above algorithm, each minECG (or pseudo-minECG) was tagged as either stable or unstable based on its P_m value. To test the 1^st hypothesis, the number of unstable minECGs (N_us) was counted for each patient and was compared between the VT group and the control group using a paired t-test. The frequency of PA (fpA) was also determined for each patient, and the relationship between N_us and f_PA was determined by calculating the correlation coefficient between these two variables for each group. We also compared fpA between the two groups with a paired t-test. The significance level of all these tests was 0.05. The second hypothesis was tested by comparing the P_m indices of pseudo- minECGs with PAs of different DOP.

Results

ARX modeling

[0065] An ARX model (Equation 1) was constructed for each minECG (or pseudo- minECG) and the value of each QTI in the minECG predicted. Figures 7A-7C show the predicted QTI dynamics for a single minECG, compared to the same dynamics extracted from the patient ECG recording. Figure 7A demonstrates that an accurate prediction of QTI dynamics was achieved for ^_«=24. Figure 7B shows the predicted and the measured QTI dynamics for =3. The dependence of QTI prediction accuracy on M is presented in Figure 7C for the same minECG. The mean value of Μ^_χ of the VT group was 37.8+8.8, which was significantly (p<0.00001) different from that of the control group (32.1+8.16).

QTI instability in VT group vs. control

[0066] The value of N_us was calculated for each patient in both groups, and is presented in Table 1. A paired t-test of the results in Table 1 showed that the mean N_us value in the VT group (4.4+3.6) was significantly larger (p=0.003) than that in the control group 2240- 293156

(1.6+2.1) (Figure 8A), which means that more minECGs became unstable before VT onset. We also found that the mean f_PA of the VT group was higher, 12.9+17.96 beat/minute, than that of the control group, 3.3+8.9 beat/minute (p=0.029, Figure 8B), indicating that the frequency of PAs increased before VT onset. Finally, we documented that P_m values were always <1 in all PA-free minECGs, i.e., QTI dynamics in these minECGs was always stable. The correlation coefficient between N_us and f_PA was found to be 0.69 (p<0.05) in the VT group and 0.85 (p<0.01) in the control group.

[0067] Table 1: The number of unstable minECG in the VT group compared to those in the control group.

QTI instability in pseudo-minECGs with PAs of different DOP

[0068] Under the PA-free pacing train (DOP of 0ms), the QTI dynamics in the pseudo-minECG was stable (as it was in the original minECG from which the pacing train was extracted). For the modified pacing trains with non-zero DOP (DOP of 150ms, 200ms, and 220ms), the values of P_m of the corresponding pseudo-minECGs were 0.97, 1.001, and 1.02, respectively. Unstable QTI dynamics were thus observed for DOP of 200ms and 220ms. Clearly, increasing DOP revealed larger instability in QTI dynamics.

Discussion

[0069] This example presents an embodiment of a novel method for detecting instability in QTI dynamics from the ECG recording. By using this methodology, we were able to assess the contribution of unstable QTI dynamics to VT onset, and to explore the relationship between PA and QTI instability. The results of this study revealed increased frequency of PA and QTI instability before VT onset, and established that QTI instability is positively correlated with the DOP of PA. These findings shed light on the mechanisms of 2240- 293156 arrhythmogenesis, and indicate a possibility for the development of an index that predicts risk of arrhythmias from the clinical ECG.

[0070] This study revealed important contribution of PA and QTI instability to the onset of VT. It has been reported that unstable APD dynamics following PA is responsible for the VT onset. Researchers have hypothesized that unstable APD dynamics, caused by PA and large (>1) APD restitution slope, could lead to heterogeneous distribution of DI, and consequentially to the induction of conduction block, reentry, and VT. 1-^"3 However, this hypothesis has not been confirmed in the clinic. The measurement of APD restitution slope requires constant pacing protocols to eliminate short-term memory, while in the clinical setting, heart rhythm before arrhythmia onset is typically non-constant, and the contribution of short-term memory to QTI instability cannot be ignored. There is currently no reliable way to assess QTI instability from clinical recordings since short-term memory is present. The present study filled this void with a novel methodology that assesses QTI instability, with short-term memory being included. Applying this methodology to clinical ECG recorded from AMI patients, we identified unstable QTI dynamics before VT onset.

[0071] The contribution of PA and QTI instability to VT onset, as discussed in the above sections, is based on the assessment of QTI instability. In this study we represented QTI dynamics as an ARX model, an approach that has been used previously to determine the contribution of restitution and short-term memory to the value of APD.¹⁶' ²⁹ An ARX model

M

(Equation 1) is composed of an autoregressive part (the x QTI_n__{ part of Equation 1) and i=l

M

an exogenous part (the x RRI_n__{ part of Equation 1). A novel aspect of the present i=l

methodology is the instability analysis of the ARX model.

[0072] Previous restitution research has reported that the restitution slope depends on the value of the preceding DI (or TQI). That is, restitution slope increases when DI (or TQI) decreases, and large (>1) restitution slope was only observed for small DI (or TQI) ranges. Performing the instability analysis of the ARX model, we observed a similar relationship between the DOP of PA and QTI instability. We found that unstable QTI dynamics was only 2240- 293156 observed when PA was presented, and the increased DOP of PA increased the instability of QTI dynamics.

[0073] The advantage of the present algorithm, compared against restitution, is that it is designed to accommodate short-term memory, while restitution is not. To demonstrate this advantage, a QTI restitution curve was constructed from QTI and TQI values of one minECG of the VT group (Figure 9). PAs were presented in this minECG, and unstable QTI dynamics was identified with the present methodology. From Figure 9, we can see that the QTI restitutions in the small TQI range (TQI<=400ms) and the large TQI range (TQI>400ms) are very different, and cannot be properly fitted with a single restitution curve. Instead, we fitted two exponential curves in these two ranges. In the small TQI range, the QTI restitution slope was negative; in the large TQI range, the slope is smaller than 1. Both slopes indicate stable QTI dynamics. Note that the curve fit presented in Figure 9 was not possible for many other minECGs because large scatter in QTI restitution data points, indicative of strong short-term memory, exists in those minECGs.

Clinical significance

[0074] The results of this study can be applied in both clinical monitoring for arrhythmia events and arrhythmia risk stratification. An important finding of the present study is that both f_PA and Nus increased before VT onset. We also found that QTI dynamics instability was dependent on the DOP of PA. These findings open a new way to predict the initiation of VT by monitoring fpA, the DOP of PA, and Nus in clinical ECG. We also want to emphasize that, because all the patients enrolled in this study had CAD, the positive correlation between f_PA and Nus reported for this example may not apply to the healthy heart. If this is proven to be the case, a positive correlation between fpA and Nus in the ECG would indicate arrhythmic risk, and could be used as a risk stratification index.

[0075] The references cited for this example follow, in numerical order.

[0076] REFERENCES

1. Gilmour F, Gelzer AR, Otani NF. Cardiac electrical dynamics: maximizing dynamical

heterogeneity. 2007;40(6):S51-S55. 2240- 293156

2. Fox JJ, Riccio ML, Drury P, Werthman A, Gilmour RF. Dynamic mechanism for conduction block in heart tissue. New Journal of Physics. 2003;5: 101.101-101.104.

3. Otani NF. Theory of action potential wave block at-a-distance in the heart. PHYSICAL REVIEW E. 2007;75:021910-021911 - 021910-021917.

4. Laurita KR, Girouard SD, Akar FG, Rosenbaum DS. Modulated Dispersion Explains Changes in Arrhythmia Vulnerability During Premature Stimulation of the Heart. Circulation.

1998;98:2774-2780.

5. Laurita KR, Girouard SD, Rosenbaum DS. Modulation of ventricular repolarization by a

premature stimulus. Role of epicardial dispersion of repolarization kinetics demonstrated by optical mapping of the intact guinea pig heart. Circulation Research. 1996;79(3):493-503.

6. Narayan SM, Lindsay BD, Smith JM. Demonstration of the Proarrhythmic Preconditioning of Single Premature Extrastimuli by Use of the Magnitude, Phase, and Distribution of

Repolarization Alternans. Circulation. 1999;100:1887-1893.

7. Nolasco JB, Dahlen RW. A graphic method for the study of alternation in cardiac action

potentials. Journal Of Applied Physiology. 1968;25(2): 191-196.

8. Roller ML, Riccio ML, Gilmour RF. Dynamic restitution of action potential duration during electrical alternans and ventricular fibrillation. Am J Physiol. 1998;44:H1635-H1642.

9. Garfinkel A, Kim Y, Voroshilovsky O, Qu Z, Kil J, Lee M, Karagueuzian H, Weiss J, Chen P.

Preventing ventricular fibrillation by flattening cardiac restitution. Proc Natl Acad Sci U S A. . 2000;97:6061-6066.

10. Chen X, Fenton FH, Gray RA. Head-tail interactions in numerical simulations of reentry in a ring of cardiac tissue. Heart Rhythm: The Official Journal Of The Heart Rhythm Society.

2005;2(9): 1038-1046.

11. Watanabe MA, Fenton FH, Evans SJ, Hastings HM, Karma A. Mechanisms for discordant alternans. Journal Of Cardiovascular Electrophysiology. 2001;12(2):196-206.

12. Qu Z, Garfinkel A, Chen PS, Weiss JN. Mechanisms of discordant alternans and induction of reentry in simulated cardiac tissue. Circulation. 2000;102(14):1664-1670.

13. Roller ML, Maier SKG, Gelzer AR, Bauer WR, Meesmann M, Robert F. Gilmour J. Altered Dynamics of Action Potential Restitution and Alternans in Humans With Structural Heart Disease. Circulation. 2005;112:1542-1548.

14. Selvaraj RJ, Picton P, Nanthakumar K, Chauhan VS. Steeper restitution slopes across right ventricular endocardium in patients with cardiomyopathy at high risk of ventricular arrhythmias. Am J Physiol Heart Circ Physiol 2007;292:H1262-H1268.

15. SHIFERAW Y, QU Z, GARFINKEL A, KARMA A, WEISS JN. Nonlinear Dynamics of Paced Cardiac Cells. Annals of the New York Academy of Sciences. 2006;1080(l):376-394.

16. Huang J, Zhou X, Smith WM, Ideker RE. Restitution properties during ventricular fibrillation in the in situ swine heart. Circulation. 2004;110(20):3161-3167.

17. Gilmour RF, Otani NF, Watanabe MA. Memory and complex dynamics in cardiac Purkinje fibers. American Journal of Physiology. 1997;272(4):H1826.

18. Otani NF, Gilmour RF. Memory Models for the Electrical Properties of Local Cardiac Systems.

/. Theor. Biol. 1997;187:409-436.

19. Yamauchi S, Yamaki M, Watanabe T, Yuuki K, Kubota I, Tomoike H. Restitution properties and occurrence of ventricular arrhythmia in LQT2 type of long QT syndrome. Journal Of Cardiovascular Electrophysiology. 2002;13(9):910-914.

20. Fossa AA, Wisialowski T, Crimin K, Wolfgang E, Couderc J-P, Hinterseer M, Kaab S, Zareba W, Badilini F, Sarapa N. Analyses of Dynamic Beat-to-Beat QT-TQ Interval (ECG Restitution) Changes in Humans under Normal Sinus Rhythm and Prior to an Event of Torsades de Pointes during QT Prolongation Caused by Sotalol. Annals Of Noninvasive Electrocardiology: The Official Journal Of The International Society For Holier And Noninvasive Electrocardiology, Inc. 2007;12(4):338 - 348.

21. Sachdev M, Fetics BJ, Lai S, Dalai D, Insel J, Berger RD. Failure in Short-term Prediction of Ventricular Tachycardia and Ventricular Fibrillation from Continuous ECG in Intensive Care Unit Patients /. Electrocardiology. 2010;(in press). 2240- 293156

22. Laguna P, Jane R, Caminal P. Automatic Detection of Wave Boundaries in Multilead ECG

Signals: Validation with the CSE Database. Computers and Biomedical Research. 1994;27(1):45- 60.

23. Huikuri HV, Valkama JO, Airaksinen KE, Seppanen T, Kessler KM, Takkunen JT, Myerburg RJ. Frequency domain measures of heart rate variability before the onset of nonsustained and sustained ventricular tachycardia in patients with coronary artery disease. Circulation.

1993;87(4):1220-1228.

24. Vadakkumpadan F, Rantner L, Tice B, Boyle P, Prassl A, E.Vigmond, Plank G, Trayanova N.

Image-based models of cardiac structure with applications in arrhythmia and defibrillation studies. Journal of Electrocardiology. 2009;42(2):157.el51-157.ell0.

25. Prassl A, Kickinger F, Ahammer H, Grau V, Schneider J, Hofer E, Vigmond E, Trayanova N, Plank G. Automatically Generated, Anatomically Accurate Meshes for Cardiac

Electrophysiology Problems. IEEE-TBME. 2009;56:1318 - 1330.

26. ten Tusscher KH, Noble D, Noble PJ, Panfilov AV. A model for human ventricular tissue.

American Journal Of Physiology Heart And Circulatory Physiology. 2004;286(4):H1573-1589.

27. Maharaj T, Blake R, Trayanova N, Gavaghan D, Rodriguez B. The role of transmural

ventricular heterogeneities in cardiac vulnerability to electric shocks. Progress In Biophysics And Molecular Biology. 2008;96(l-3):321-338.

28. Drouin E, Charpentier F, Gauthier C, Laurent K, Le Marec H. Electrophysiologic

characteristics of cells spanning the left ventricular wall of human heart: evidence for presence of M cells. J Am Coll Cardiol. 1995;26(1): 185-192.

29. Toal SC, Farid TA, Selvaraj R, Chauhan VS, Masse S, Ivanov J, Harris L, Downar E, Franz MR, Nanthakumar K. Short-Term Memory and Restitution During Ventricular Fibrillation in Human Hearts An In Vivo Study Circulation: Arrhythmia and Electrophysiology. 2009;2:562-570.

30. Nelles O. Nonlinear System Identification. Springer-Verlag Berlin Heidelberg New York. 2001:1.

31. Steiglitz K, McBRide LE. A Technique for the Identification of Linear Systems. IEEE Trans.

Automatic Control. 1965;AC-10:461-464.

32. Lathi BP. LINEAR SYSTEMS AND SIGNALS. Berkeley-Cambridge Press. 1992:227.

33. Pastore JM, Girouard SD, Laurita KR, Akar FG, Rosenbaum DS. Mechanism linking T-wave alternans to the genesis of cardiac fibrillation. Circulation. 1999;99(10): 1385-1394.

34. Pruvot EJ, Katra RP, Rosenbaum DS, Laurita KR. Role of Calcium Cycling Versus Restitution in the Mechanism of Repolarization Alternans. Circulation Research. 2004;94: 1083: 1090.

35. Chinushi M, Restivo M, Caref E, El-Sherif N. Electrophysiological basis of arrhythmogenicity of QT/T alternans in the long QT syndrome: tridimensional analysis of the kinetics of cardiac repolarization. Circ Res. 1998;83:614 -628.

36. Rosenbaum DS, Jackson LE, Smith JM, Garan H, Ruskin JN, Cohen RJ. Electrical Alternans and Vulnerability to Ventricular Arrhythmias. N Engl J Med. 1994;330:235-241.

37. Fox JJ, McHarg JL, Gilmour RF. Ionic mechanism of electrical alternans. Am J Physiol Heart Circ Physiol. 2002;282:H516-H530.

EXAMPLE 2

[0077] We have developed a novel algorithm to detect instability in QT interval

(QTI) dynamics from clinical ECG recordings or ICD electrograms, as presented in Example 1. We demonstrated that QTI dynamics instability precedes the onset of ventricular tachycardia (VT) in patients with structural disease. We also demonstrated that the onset of VT is associated with increased frequency of premature activations (PA). In this example, 2240- 293156 using the developed algorithm to detect QTI dynamics instability, we constructed an arrhythmia risk stratification index. This index determines the dependence of QTI dynamics instability on the presence of PAs.

[0078] This example shows that the dependence of the instability in QTI dynamics on the presence of PA can be used to stratify arrhythmia risk in patients with structural heart disease.

[0079] The patient population had structural heart disease with an ICD implanted to provide an ICD recording, i.e., an EGM. The follow up of each patient terminated when the following endpoint events occurred:

1. The patient received appropriate ICD therapies (shock or pacing) when VT/VF was detected; or

2. Death.

[0080] For each patient, the risk stratification index needed assessment of the instability in QTI dynamics. The ECG or the electrogram trace from the ICD of each patient was divided into 1 -minute long traces termed minECGs. The algorithm was applied to each minECG to detect whether QTI dynamics was unstable. The number of minECGs with unstable QTI dynamics was Nus- The number of minECGs with PAs present was N_PA-

[0081] The arrhythmia risk stratification index, termed QTI instability index (QTII), was defined as

[0082] Our rationale is to consider N_PA as the perturbation of a system, and Nus as the response of the system. QTII is the sensitivity of the system to the perturbation, that is, the sensitivity of Nus to N_PA- QTII is this always less than 1.

[0083] To evaluate the prediction value of QTII, we used the following approaches: 2240- 293156

1. Constructed Receiver Operating Characteristic (ROC) curve (sensitivity vs. 1- specificity for different QTII thresholds); and

2. Conducted Survival analysis (Kaplan-Meier and Cox).

[0084] Intracardiac electrocardiograms (EGMs) were recorded at rest in 44 patients with implanted ICDs. Patients were followed up for 29+16 months. During the follow up period, 11 patients received proper ICD therapies, and 8 patients died.

[0085] Table 2: Retrospective Analysis

[0086] In the above table, the percentage of unstable minECGs was calculated over the total number of minECGs in each patient. From Table 2, one can see that the high-risk group (patients with endpoint events) has a larger QTII (significant) and more unstable minECGs (significant).

[0087] The sensitivity, specificity, relative risk (RR), and efficiency of the prediction for several example QTII values are listed in Table 3. The relationship between the sensitivity and specificity of the prediction for the QTII values used here is represented by the ROC curve (Figure 10).

[0088] Table 3: Prospective Analysis

2240- 293156

[0089] The area under the ROC curve was 0.76 (Figure 10). For different QTII values

(1>=QTII>=0.5), the relative risk ranges from 2.5 to 3.3; the hazard ratio (Cox survival analysis) ranges from 3.4 to 5.5 (p<0.05); the Log rank p value (Kaplan Meier test, Figure 11) ranges from 0.0003 to 0.01. These results demonstrate that QTII can be used in the risk stratification of arrhythmia according to an embodiment of the current invention.

[0090] The embodiments illustrated and discussed in this specification are intended only to teach those skilled in the art the best way known to the inventors to make and use the invention. In describing embodiments of the invention, specific terminology is employed for the sake of clarity. However, the invention is not intended to be limited to the specific terminology so selected. The above-described embodiments of the invention may be modified or varied, without departing from the invention, as appreciated by those skilled in the art in light of the above teachings. It is therefore to be understood that, within the scope of the claims and their equivalents, the invention may be practiced otherwise than as specifically described.

Claims

2240- 293156 WE CLAIM:

1. A method of predicting ventricular arrhythmias, comprising:

receiving an electrical signal from a subject's heart for a plurality of heart beats; identifying characteristic intervals and heart beat durations of the electrical signal corresponding to each of the plurality of heart beats to provide a plurality of characteristic intervals with corresponding heart beat durations;

representing dynamics of said plurality of characteristic intervals as a function of a plurality of preceding characteristic intervals and durations of corresponding heart beats over a chosen period time;

assessing a stability of said function over said chosen period of time; and

predicting ventricular arrhythmias based on detected instabilities in the dynamics of said characteristic intervals.

2. A method of predicting ventricular arrhythmias according to claim 1, wherein said receiving an electrical signal is at least one of receiving an ECG or ICD electrogram signal.

3. A method of predicting ventricular arrhythmias according to claim 2, wherein said characteristic intervals are QT intervals of said at least one of said ECG or ICD electrogram signal.

4. A method of predicting ventricular arrhythmias according to claim 3, wherein said representing the dynamics of said plurality of characteristic intervals includes determining parameters of a QT interval dynamics model based on said QT intervals and on RR intervals.

5. A method of predicting ventricular arrhythmias according to claim 4, wherein said QT interval dynamics model is an ARX model. 2240- 293156

6. A method of predicting ventricular arrhythmias according to claim 4, wherein said determining parameters of a QT interval dynamics model based on said QT intervals and RR intervals is determining parameters a; and bj in the following equation

M M

QTI_n =∑ a, x QTI^ +∑ b_t x RRI^

i=l i=l

where M is a number of heart beats included in the determining and n is a heart beat number.

7. A method of predicting ventricular arrhythmias according to claim 4, wherein said assessing the stability of said dynamics of said characteristic intervals comprises determining the stability of said model of said QT interval dynamics.

8. A method of predicting ventricular arrhythmias according to claim 1, further comprising:

repeating said representing dynamics and said assessing a stability for a plurality of chosen periods of time; and

counting a number of said plurality chosen periods of time in which an instability is detected to obtain a number represented as Nus-

9. A method of predicting ventricular arrhythmias according to claim 8, further comprising:

counting a number of said plurality chosen periods of time in which a premature heart beat occurs to obtain a number represented by N_PA;

calculating a ratio of Nus divided by N_PA to obtain a stratification index; and predicting ventricular arrhythmias based on said stratification index.

10. A method of predicting ventricular arrhythmias according to claim 3, further comprising:

counting a number of said plurality chosen periods of time in which an instability is detected to obtain a number represented as Nus- 2240- 293156

11. A method of predicting ventricular arrhythmias according to claim 10, further comprising:

12. A medical device comprising a system for predicting ventricular arrhythmias, said system comprising a data processor configured to:

receive an electrical signal from a subject's heart for a plurality of heart beats;

identify characteristic intervals and heart beat durations of the electrical signal corresponding to each of the plurality of heart beats to provide a plurality of characteristic intervals with corresponding heart beat durations;

represent the dynamics of said plurality of characteristic intervals as a function of a plurality of preceding characteristic intervals and durations of the corresponding heart beats over a chosen period of time;

assess a stability of said function over said chosen period of time; and

predict ventricular arrhythmias based on detected instabilities in the dynamics of said characteristic intervals.

13. A medical device according to claim 12, wherein said electrical signal is at least one of an ECG or ICD electrogram signal.

14. A medical device according to claim 13, wherein said characteristic interval is a QT interval of said ECG or ICD electrogram signal.

15. A medical device according to claim 14, wherein said model of the dynamics of said plurality of characteristic intervals includes parameters determined for a QT interval model based on said plurality of QT and RR intervals. 2240- 293156

16. A medical device according to claim 15, wherein said model of the dynamics of said plurality of characteristic intervals is an ARX model.

17. A medical device according to claim 15, wherein said model of QT interval dynamics can be represented by the equation

M M

QTI_n =∑ a, x QTI^ +∑ b_t x RRI^

i=l i=l

where M is a number of heart beats included in the determining parameters a; and b; and n is a heart beat number.

18. A medical device according to claim 15, wherein said data processor is configured to determine stability of said dynamics of said characteristic intervals comprising determining the stability of said model of said QT interval dynamics..

19. A medical device according to claim 12, wherein said data processor is further configured to:

repeat said representing dynamics and said assessing a stability for a plurality of chosen periods of time; and

count a number of said plurality chosen periods of time in which an instability is detected to obtain a number represented as Nus-

20. A medical device according to claim 19, wherein said data processor is further configured to

count a number of said plurality chosen periods of time in which a premature heart beat occurs to obtain a number represented by N_PA;

calculate a ratio of Nus divided by N_PA to obtain a stratification index; and predict ventricular arrhythmias based on said stratification index. 2240- 293156

21. A medical device according to claim 14, wherein said data processor is further configured to:

22. A medical device according to claim 21, wherein said data processor is further configured to:

calculate a ratio of Nus divided by N_PA to obtain a stratification index; and predict ventricular arrhythmias based on said stratification index.

23. A medical device according to any one of claims 12 to 22, wherein said medical device is at least one of a Holter monitor, a treadmill stress test device, an implantable cardioverter-defibrillator, or an ECG-based device.

24. A computer readable medium comprising stored executable instructions for execution by a computer, comprising executable instructions for:

assessing a stability of said function over said chosen period of time; and

predicting ventricular arrhythmias based on detected instabilities in the dynamics of said characteristic intervals. 2240- 293156

25. A computer readable medium according to claim 24, wherein said receiving an electrical signal is at least one of receiving an ECG or ICD electrigram signal.

26. A computer readable medium according to claim 25, wherein said characteristic interval is a QT interval of said ECG or ICD electrogram signal.

27. A computer readable medium according to claim 26, wherein said wherein said representing the dynamics of said plurality of characteristic intervals includes determining parameters of a QT interval dynamics model based on said QT intervals and on RR intervals.

28. A computer readable medium according to claim 27, wherein said wherein said QT interval dynamics model is an ARX model.

29. A computer readable medium according to claim 27, wherein said determining parameters of a QT interval dynamics model based on said QT intervals and RR intervals is determining parameters a; and bj in the following equation

M M

QTI_n =∑ a, x QTI^ +∑ b_t x RRI^

;=i i=i

30. A computer readable medium according to claim 27, wherein said assessing the stability of said dynamics of said characteristic intervals comprises determining the stability of said model of said QT interval dynamics.

31. A computer readable medium according to claim 27, wherein said determining the stability of said dynamics of said characteristic intervals comprises determining the stability of said model of said QT interval dynamics.

32. A computer readable medium according to claim 24, further comprising further comprising: 2240- 293156 repeating said representing dynamics and said assessing a stability for a plurality of chosen periods of time; and

33. A computer readable medium according to claim 32, further :

calculate a ratio of Nus divided by N_PA to obtain a stratification index; and

predicting ventricular arrhythmias based on said stratification index.

34. A computer readable medium according to claim 27, further comprising:

35. A computer readable medium according to claim 34, further comprising:

calculate a ratio of Nus divided by N_PA to obtain a stratification index; and

predicting ventricular arrhythmias based on said stratification index.